SMBlog — 6 December 2013

Alternate Universes: Academic Publishing in Computer Science vs. Law

I (and my co-authors) have recently had two papers accepted to law reviews, "When enough is enough: Location tracking, mosaic theory, and machine learning", and "Lawful hacking: Using existing vulnerabilities for wiretapping on the Internet". In the process, I learned something about the world of legal academic publishing. It’s about as different from my more familiar world of computer science publishing as can be. With the thought of amusing everyone, both on the legal and CS sides, I thought I’d make a list of differences…

Computer Science	Law Reviews
In CS (and I believe in all other STEM fields), multiple submission is strictly forbidden. That is, you cannot simultaneously submit substantially the same paper to more than one venue. If the program chairs discover that you have done so, it is grounds for immediate summary rejection of the paper from all such venues, with no appeals considered. I’ve seen it happen—and yes, chairs do check for this.	In legal academe, not only is multiple submission accepted, it is the normal, right, and proper way to do things. You’re supposed to submit to many—dozens, perhaps—law reviews simultaneously. If your paper is accepted by one, you withdraw it from any lower-ranked journals and ask the editors of higher-ranked journals to expedite their reviews. ("Expedited" can mean "within a week", perhaps less, because that’s how long you may have to accept the earlier offer.) If you receive another acceptance from a publication that is higher-ranked still, you repeat the process until you’re either satisfied or you hit the deadline for accepting an offer to publish.
Every submission is independent. If a paper is rejected from one venue, you look for the next, and use its submission system. Submissions, though, are free.	There are two centralized sites for law reviews. On these sites, you check the publications you’re interested in; they do the rest. Partly to pay for this service, and partly to impose a limit on how many sites you submit to, there is a fee per publication. It’s not a large fee, just a few dollars, but if you submit to 50 venues, it adds up. (Many law schools pay for blanket submission licenses. That’s also foreign to CS; were there such charges, each prof would pay out of his or her grants.)
CS and all other sciences worship peer review. That is, submissions are judged by professionals in the field: professors, high-end practitioners, etc. Being on a conference program committee is an honor, albeit a very time-consuming one. Some committes will have a student member or two; generally, this is reserved for very senior graduate students who have a good publication track record of their own. The rationale is simple: it takes a good researcher to know what is really new and interesting; students are presumed to have too little experience (and often too little historical context).	Most law reviews are edited by law students. Much of law school is about teaching critical thinking (a philosphy I’d like to see more of in CS…); good law students are presumed to be able to recognize sound arguments and good scholarship.
CS conferences are scattered throughout the year; you can submit to whichver is convenient and appropriate for your topic. The major security conferences even arrange their notification and submission deadlines so that you (just barely) have time to revise a rejected paper in time to submit to the next conference.	Because law reviews are edited by students, there are two publication windows, August and February. Virtually all law reviews follow that calendar. Furthermore, there’s no hard submission deadline; rather, journals accept papers unti the issue is full.
For most CS conferences (and certainly all of the major security conferences), submissions are anonymous. This is done to avoid bias, be it based on personal feelings, the reputation of the authors, or (and I’ve heard this stated explicitly by program chairs) the authors’ gender.	Perhaps because law students are inexperienced, submissions are not only not anonymous, they’re accompanied by biographies and even full CVs.
Reviewers of CS papers explain their decisions, pro or con, and provide detailed feedback. You can’t just reject a paper you don’t like; rather, you have to explain exactly what’s wrong. While some reviewers don’t do a good job (see this set of instructions on how to do better), authors frequently receive very useful feedback. (Admittedly, sometimes the feedback is implicit: reviewer #2 either didn’t read or didn’t understand the paper….)	Law review rejections are terse: "thank you very much, but we’re not interested". Amusingly enough, that means that CS rejections are more like legal opinions: reviewers have to justify their decisions. I suppose the law review model is more like the Supreme Court denying certiorari.
Editing of CS papers is generally cursory. There may be a few comments about getting the paper reviewed by a native (English) speaker, or some serious problems with content that must be changed to the satisfaction of a "shepherd".	The level of editing and attention to detail by law review editors is nothing short of amazing. One of my papers literally had 4,917 changes made by the first-round editors: 1,943 insertions, 1,774 deletions, 2 moves, and 1,088 formatting changes. In addition, there were 110 comments.
From the time of acceptance to the time that cameara-ready copy is due, you have a long time to make changes: 8 weeks for last year’s Usenix Security conference.	Those 4,900 changes and110 comments? We had to turn them around in 8 days.
Most computer scientists write their papers in LaTeX. Submissions are always in PDF.	Law reviews generally want Microsoft Word documents or perhaps something trivially convertible to Word, such as RTF. (Yes, you’re welcome to use something like LibreOffice if you wish.) Some (though by no means all) will apparently take PDF. To me, this was one of the more painful parts of the process…
One of the nice things about LaTeX is bibtex, a bibliographic citation formatter. If you have a good bibligraphic database (my personal one has over 1,000 entries), you just write \cite{cheswick.bellovin.ea:firewalls} and the software does the rest. (If you’ve ever looked at my papers web page, you’ll see that each entry has a bib link; clicking on it gives a bibtex database entry for that publication.) Journals generally have a preferred format, but this is a minor matter since bibtex handles that for you.	In legal writing, citations must be written in Blue Book format. This mirrors court practice; many courts require it in legal filings. However, not only is this format required, the specification is copyrighted, the copyright is enforced, and software developers are denied permission to write code to automate the task. (Very many of the ~5K edits were to fix our citation formatting.)
While academic convention does require proper citations to the academic literature, most computer scientists are fairly restrained about this. Actual footnotes are comparatively rare; citations are inline.	Lawyers love footnotes. Apart from citations to other articles, laws, court rulings, etc. (and these are full of strange italicized words like supra and infra), footnotes are used for digressions, longer explanations of issues, etc. Every factual statement must have a footnote giving a source for it. I’d write here that a typical law review article is 25% footnotes by weight (contents may have settled during shipping), but that clause would require two footnotes, one for the assertion and one for the joke. By the way, make sure you give correct page numbers for the specific facts you’re relying on; the editors will check them.
In CS, most important papers are published in conference proceedings, not journals. (Other sciences differ.) These submissions have a length limit, typically about 15 pages or so.	Law reviews sometimes talk about limits, but these are lightly enforced, and the typical minimum size is far longer than CS’s maximum. Let’s put it like this: the shorter of my two accepted papers is about 50 pages; the other is considerably longer. Lawyers and law professors say they want papers to be shorter, but that doesn’t seem to be happening in practice.
Everyone wants their work to have impact. In CS, the h-index is commonly used; without going into details, it measures how often a paper is cited.	Lawyers also care about citations, but even a single citation can make a paper’s (and hence its author’s) reputation—if that citation is in an important court opinion.

Neither publishing model is perfect. Many computer scientists think that the stress on conferences is harmful; I’ve heard law professors debate the wisdom of student editors. Some things, like single versus multiple submissions, are matters of custom and taste. (If you’re a lawyer, note that this difference is the one that most perplexes my colleagues. In fact, I’m not sure I believe it even now.)

Still, it’s good to visit other planets occasionally, even if it’s just academically.

https://www.cs.columbia.edu/~smb/blog/2013-12/2013-12-06.html